Multi-Agent Discovery Prompt for Feature Forge

Multi-Agent Discovery Prompt for Feature Forge

Purpose: Use this prompt to trigger parallel skill-invoked discovery across multiple domains before starting a Feature Forge specification workshop. The collected findings become input context for the Feature Forge interview, replacing guesswork with concrete technical evidence.

Prompt

I need to define a new feature: [FEATURE DESCRIPTION].

Before we start the Feature Forge specification workshop, run parallel discovery
across the relevant domains to build technical context. Launch the following
Task subagents concurrently — each should invoke its respective skill and return
a focused summary:

1. **Architecture Discovery** — Launch a Task subagent (general-purpose) that invokes
   the `architecture-designer` skill to:
   - Identify which system components this feature touches
   - Map integration points and data flows
   - Flag architectural constraints or trade-offs
   - Recommend patterns that fit the existing system

2. **Security Discovery** — Launch a Task subagent (general-purpose) that invokes
   the `security-reviewer` skill to:
   - Identify authentication and authorization requirements
   - Flag data sensitivity concerns (PII, PCI, GDPR)
   - Surface OWASP-relevant risks for this feature type
   - Recommend security patterns and controls

3. **Codebase Discovery** — Launch a Task subagent (Explore) to:
   - Search for existing patterns, components, or utilities related to this feature
   - Identify code conventions and abstractions already in use
   - Find similar features that can serve as implementation templates
   - Note relevant test patterns and coverage expectations

4. **API/Integration Discovery** — Launch a Task subagent (general-purpose) that
   invokes the `api-designer` skill to:
   - Propose API surface area (endpoints, methods, payloads)
   - Identify existing API patterns to follow for consistency
   - Flag external service dependencies and rate limits
   - Recommend versioning and backwards-compatibility approach

Wait for all subagents to complete, then:

1. **Synthesize** — Combine the findings into a structured discovery summary with
   sections for: Architecture, Security, Codebase Patterns, and API Surface.
2. **Identify Decisions** — List the key decisions that emerged from discovery
   (e.g., "sync vs async processing", "new table vs extend existing").
   Present these as `AskUserQuestions` with structured options.
3. **Launch Feature Forge** — With the discovery context and user decisions in hand,
   invoke the `feature-forge` skill to begin the specification workshop. The interview
   should reference discovery findings rather than re-asking questions the subagents
   already answered.

Usage

Basic Invocation

Replace [FEATURE DESCRIPTION] with your feature summary:

I need to define a new feature: user-facing data export that supports CSV and JSON
formats with scheduled recurring exports.

Before we start the Feature Forge specification workshop, run parallel discovery...

Customizing Discovery Agents

Not every feature needs all four discovery tracks. Tailor the subagent list:

Expected Output Flow

User provides feature description
    │
    ├─→ [Parallel] Architecture Discovery ──→ Component map, constraints
    ├─→ [Parallel] Security Discovery ──────→ Auth reqs, data sensitivity
    ├─→ [Parallel] Codebase Discovery ──────→ Existing patterns, templates
    └─→ [Parallel] API Discovery ───────────→ Endpoint design, integrations
         │
         ▼
    Synthesized Discovery Summary
         │
         ▼
    AskUserQuestions: Key decisions from discovery
         │
         ▼
    Feature Forge Interview (informed by discovery context)
         │
         ▼
    EARS Specification Document

Design Rationale

This prompt applies the ReAct + Chain-of-Thought pattern from prompt engineering:

• ReAct: Each subagent acts as a reasoning-action step — it invokes a skill (action), collects findings (observation), then feeds results forward.
• Chain-of-Thought: The synthesis step forces explicit reasoning about how discovery findings connect before entering the Feature Forge interview.
• Structured Elicitation: Key decisions surface as AskUserQuestions with options derived from technical evidence, not assumptions.

The parallel execution pattern reduces total discovery latency — all four agents run concurrently rather than sequentially asking the user about each domain.